SIP Strict versus Loose Routing

The way how record routing works has evolved. Record routing according to RFC2543 rewrote the Request-URIii. That means the Request-URI always
contained URI of the next hop (which can be either next proxy server which inserted Record-Route header field or destination user agent). Because of that it was necessary to save the original Request-URI as the last Route header field. This approach is called strict routing.

Loose routing, as specified in RFC3261, works in a little bit different way. The Request-URI is no more overwritten, it always contains URI of the destination user agent. If there are any Route header field in a message, than the message is sent to the URI from the topmost Route header field. This is significant change--Request-URI doesn't necessarily contain URI to which the request will be sent. In fact, loose routing is very similar to IP source routing.

Because transit from strict routing to loose routing would break backwards compatibility and older user agents wouldn't work, it is necessary to make loose routing backwards compatible. The backwards compatibility unfortunately adds a lot of overhead and is often source of major problems.